OPA CPU and Memory Consumption

[birbalekka]

HI All,
We are running OPA in EKS as an container. While checking its CPU and memory usages , we see that the cpu doesnt go beyond 1.3 ( allocated 4 ) . Memory - 30 mb ( allocated 2 gb) .
However even with this configuration we are seeing high response times intermittently of upto 13 secs.

Is OPA getting restricted with CPU or memory ?
How can we increase it ?

arguments used during deployment

args:
- "run"
- "--ignore=.*" # exclude hidden dirs created by Kubernetes
- "--server"
- "--tls-cert-file"
- "{{ .Values.tlsCertFile }}"
- "--tls-private-key-file"
- "{{ .Values.tlsPrivateKeyFile }}"
- "--addr"
- "{{ .Values.httpsAddr }}"
- "--addr"
- "{{ .Values.httpAddr }}"
- "--log-level"
- "info"

[humbertoc-silva]

Hi @birbalekka,

What about your policies? Are they CPU-bound or IO/bound?
Maybe you are using, for example, the built-int http.send to fetch some data to use on policy evaluation and this operation is taking a lot o time. If this is your case, try substituting the http.send results with some static in-memory data to see if the latency problem disappears.

[anderseknert]

All good questions! In addition to your configuration @birbalekka , I'd suggest adding decision logging to OPA, so that you may get a better idea of what kinds of input would cause such am extreme delay.